Estimation device, estimation method, and estimation program

Abstract

To provide an estimating device, an estimating method, and an estimating program that make it possible to estimate an activity position of a target user with less information.SOLUTION: An estimating device 10 includes: a first position distribution generating unit 11 that generates a first position distribution of a target user based on account information of the target user in social media; a second position distribution generating unit 12 that generates, based on account information of a friend having a relationship with the target user in the social media, a second position distribution of the friend; and an estimating unit 13 that estimates an activity position of the target user based on the generated first position distribution and the generated second position distribution.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an estimation device, an estimation method, and an estimation program.

Background Art

[0002] In recent years, social media such as SNS (Social Networking Service) has spread worldwide and is widely used. Since a large amount of information about users having accounts and their friends is accumulated daily in social media, methods for analyzing and utilizing this information have been studied.

[0003] As related technologies, technologies for estimating a user's position from social media information are known as in Non-Patent Documents 1 and 2. Non-Patent Document 1 discloses a method for estimating an activity area of a target user using information of friends who have exchanges in the real world (offline friends) among friends who are in a friendship relationship with the target user on social media. Non-Patent Document 2 discloses a method for estimating a user's position using the user's friendship relationship and content generated by the user including text.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in Non-Patent Document 1, it is necessary to learn in advance using a large amount of data to identify offline friends. Also, in Non-Patent Document 2, in order to estimate the position using the label propagation method which is one of the semi-supervised learning methods, it is necessary to learn in advance using a large amount of data to generate a learning model. Therefore, in related technologies such as Non-Patent Documents 1 and 2, it was necessary to prepare a large amount of information in advance to estimate the activity position of the target user.

[0006] In view of such problems, an object of the present disclosure is to provide an estimation device, an estimation method, and an estimation program capable of estimating the activity position of a target user with less information.

Means for Solving the Problems

[0007] The estimation device according to the present disclosure includes a first position distribution generation unit that generates a first position distribution of the target user based on the account information of the target user in social media, a second position distribution generation unit that generates a second position distribution of the friend based on the account information of the friend related to the target user in the social media, and an estimation unit that estimates the activity position of the target user based on the generated first position distribution and the generated second position distribution.

[0008] The estimation method according to the present disclosure generates a first position distribution of a target user based on the account information of the target user in social media, generates a second position distribution of a friend based on the account information of a friend related to the target user in the social media, and estimates the activity position of the target user based on the generated first position distribution and the generated second position distribution.

[0009] The estimation program according to the present disclosure is an estimation program for causing a computer to execute a process of generating a first position distribution of a target user based on the account information of the target user in social media, generating a second position distribution of a friend based on the account information of a friend related to the target user in the social media, and estimating the activity position of the target user based on the generated first position distribution and the generated second position distribution.

Effect of the Invention

[0010] According to the present disclosure, it is possible to provide an estimation device, an estimation method, and an estimation program capable of estimating the activity position of a target user with less information.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Modes for Carrying Out the Invention

[0012] Hereinafter, embodiments will be described with reference to the drawings. In each drawing, the same elements are denoted by the same reference numerals, and redundant explanations will be omitted as necessary.

[0013] (Overview of the Embodiment) It is possible to obtain the activity range including the place of residence of the target user from social media. On many social media platforms, it is possible to describe the user's own profile such as the username, gender, age, etc., and it is possible to set one's own place of residence, etc. as part of such a profile. Also, when posting content such as moving images and text on social media, it is possible to associate information such as the shooting location and the current location of these contents. However, although it is possible to register location information in the profile and posted content, the number of users who actually register location information is extremely small. Therefore, in the embodiment, a method for estimating location information related to the target user based on the information obtained from social media is proposed.

[0014] FIG. 1 shows an overview of the estimation device according to the embodiment. The estimation device 10 according to the embodiment is a device that estimates the activity position of the target user in the physical space using social media information. For example, the estimation device 10 can be used as a means for acquiring location information when conducting marketing such as geolocation marketing and location-based marketing that implement recommendations associated with places such as the residential area and action area of the target user. Thereby, it becomes possible to implement recommendations that do not stop at simply matching hobbies. Note that it can be used not only in marketing but also in other fields.

[0015] As shown in FIG. 1, the estimation device 10 includes a first position distribution generation unit 11, a second position distribution generation unit 12, and an estimation unit 13. The first position distribution generation unit 11 generates a first position distribution of the target user based on the account information of the target user on social media. For example, the first position distribution generation unit 11 may generate a posting distribution based on the posting information (posting location) of the target user.

[0016] The second position distribution generation unit 12 generates the second position distribution of friends based on the account information of friends related to the target user in social media. For example, the second position distribution generation unit 12 may generate a friend distribution based on the activity base information (residence information) of friends.

[0017] The estimation unit 13 estimates the activity position of the target user based on the generated first position distribution and the generated second position distribution. For example, the estimation unit 13 may estimate the activity position of the target user according to the overlap between the first position distribution and the second position distribution. Also, the first position distribution and the second position distribution may be generated by a non-parametric method such as a kernel density estimation function, and the activity position may be estimated. Either one of the first position distribution and the second position distribution may be generated by a non-parametric method. The activity position to be estimated may be an activity area, or may be an ordinary activity place (residence, workplace, store visited for purposes such as shopping and dining, and the movement route in between) that the target user visits in daily life, or may be an extraordinary activity place (tourist destination, hotel, movement route, etc. during travel or business trip) that the target user does not visit in daily life.

[0018] In this way, in the embodiment, by using the position distribution based on the account information of the target user and the position distribution based on the account information of friends, the activity position (activity area) of the target user can be estimated with less information. For example, it may be possible to estimate the activity area even when only one of the posting information of the target user or the friend information of friends is available. When two types of information are available, the activity area can be estimated more accurately by combining them. Also, by using a non-parametric method that does not require large-scale data collection, the data collection cost of social data with limitations in data collection can be reduced.

[0019] (Embodiment 1) Hereinafter, Embodiment 1 will be described with reference to the drawings. FIG. 2 shows a configuration example of the activity area estimation system according to the present embodiment. As shown in FIG. 2, the activity area estimation system 1 according to the present embodiment includes an activity area estimation device 100 and a social media system 200.

[0020] The social media system 200 is a system that provides social media services such as SNS. The social media system 200 may include a plurality of social media services. A social media service is an online service that can transmit (publish) information and communicate between a plurality of accounts (users) on the Internet (online). The social media service includes not only SNS, but also messaging services such as chat, blogs and bulletin boards (forum sites), video sharing sites and information sharing sites, social games and social bookmarks, etc.

[0021] For example, the social media system 200 includes a server on the cloud and user terminals. The server may be a social media server or a web server. The user terminal logs in with the user's account via the API (Application Programming Interface) provided by the server, inputs and browses posts, and also registers the connection of accounts such as friendship and following relationships. The social media system 200 and the activity area estimation device 100 are communicably connected via the Internet or the like.

[0022] The activity area estimation device 100 includes a post information acquisition unit 101, a post distribution generation unit 102, a friend information acquisition unit 103, a friend distribution generation unit 104, an activity area estimation unit 105, and an activity area output unit 106. Note that the configuration of each unit (block) is an example, and as long as the operations (methods) described later are possible, it may be configured with other units. Also, each unit may be provided in one device or in a plurality of devices. For example, the post information acquisition unit 101 and the post distribution generation unit 102 may be used as a first position distribution generation unit, and the friend information acquisition unit 103 and the friend distribution generation unit 104 may be used as a second position distribution generation unit.

[0023] The post information acquisition unit (target account information acquisition unit) 101 acquires the post information of the target account from the social media system 200. The post information acquisition unit 101 is also a target account identification unit that identifies the target account of the target user for whom the activity area is to be estimated. For example, the target user may be a person targeted for marketing, but may be any other person. The post information acquisition unit 101 acquires the account information (social media information) of the identified target account from the social media system 200. The account information is public information regarding the account on the social media and includes the profile information and post information of the account. The post information acquisition unit 101 may acquire the account information of a plurality of social media. The post information acquisition unit 101 may acquire it via an API or a crawler (acquisition tool) from a server that provides the social media service, or may acquire it from a database in which the account information of the social media has been stored in advance.

[0024] The post information acquisition unit 101 acquires all post information from the account information of the target account. The post information includes images, texts, etc. posted by the account (user) on the timeline, etc. The post information acquisition unit 101 extracts the post location and the post date and time from the images and texts of the acquired post information. The post location is the location where the target user posted the post information, and the post date and time is the date and time when the post information was posted. The post date and time are registered in association with the posted image or text at the time of posting. The post location is position information that can be extracted from the post information, and may be a GEO tag such as GPS (Global Positioning System) information attached to the post image, or a position specified from the reflection of landmarks, etc. in the post image. Also, not limited to images, it may be a location mentioned in the post text. The location mentioned in the post text is extracted, for example, by natural language processing of the post text. Note that the post location is an example of position information for estimating the activity location (location with a connection) of the target user from the account information of the target user, and not limited to the post location, it may also be an activity base such as the place of residence included in the profile information.

[0025] The post distribution generation unit 102 generates the post distribution (first position distribution) of the target account based on the post information of the target account. The post distribution generation unit 102 generates the post distribution of the post locations of the extracted target account. The post distribution is the distribution of the post locations (post positions) in the physical space (spatial distribution peculiar to the post positions), and is, for example, a two-dimensional geographical spatial distribution consisting of latitude and longitude coordinates. For example, the post distribution is the distribution of the post locations in units of distribution areas of a predetermined size. The granularity level of the distribution area may be an administrative division unit such as country unit, prefecture unit, city, town and village unit, or a mesh unit of a predetermined size such as 1Km×1Km, 100m×100m, 10m×10m.

[0026] The submission distribution generation unit 102 obtains a submission distribution using a predetermined distribution function. It is preferable to use a density estimation function that estimates the distribution by a non-parametric method. In this embodiment, a kernel density estimation function is used as an example of the density estimation function of the non-parametric method. In the generation (calculation) of the submission distribution, weighting may be performed on each submission information based on the submission information. For example, weighting may be performed on the submission information according to the submission date and time. Note that not limited to the distribution function, the submission distribution may be obtained by other statistical processes. For example, a submission distribution (histogram) may be generated by counting the number of submission locations included in each distribution area.

[0027] The friend information acquisition unit 103 acquires friend information of a friend account from the social media system 200. The friend information acquisition unit 103 is also a friend account identification unit that identifies the friend account of the target user. A friend account is an account that has a connection such as a friendship relationship with the target account in the social media. It may be an account of the same social media as the target user, or an account of a different social media. For example, a friend account may be an account in which a friendship relationship is registered with the target account, but may also be an account (related account) that has other connections (relationships) with the target account. For example, it may be an account with a connection of a follow relationship (following or follower), a connection by a submission (quotation such as a comment or retweet on a submission, a reaction such as "like", a mention, etc.), or an exchange history of messages. Note that a retweet is to post a comment or the like in a form that quotes a submission of another account or a submission of one's own account. A mention is to post a comment or the like that includes a specific account name.

[0028] The friend information acquisition unit 103 acquires the account information of the specified friend accounts from the social media system 200. The method of information acquisition from the social media system 200 is the same as that of the post information acquisition unit 101, and the account information is acquired by means of the API of the server or the like. The friend information acquisition unit 103 extracts friend information from the account information of all the acquired friend accounts. The friend information is location information regarding the friend accounts, for example, the place of residence (residential area) extracted from the account information. The friend information acquisition unit 103 extracts the place of residence information from the profile information included in the account information. Not limited to the place of residence, other activity bases such as the place of origin, workplace, school, etc. may be extracted. Note that the friend information is an example of location information for estimating the activity places (places with a relationship) of friends from the friend account information, and is not limited to activity bases such as the place of residence, and may also be the posting place of the post information, etc.

[0029] The friend distribution generation unit 104 generates the friend distribution (second location distribution) of the friend accounts based on the friend information (activity bases) of the friend accounts. The friend distribution generation unit 104 generates the friend distribution of the places of residence of the extracted friend accounts. The friend distribution is, like the post distribution, the distribution of the places of residence (friend locations) of friends in the physical space (the spatial distribution peculiar to the places of residence of friends). The granularity level of the distribution area of the friend distribution is the same as that of the post distribution, but may be different granularities. The friend distribution generation unit 104 obtains the friend distribution by means of a distribution function of a non-parametric method such as a kernel density estimation function, similar to the post distribution generation unit 102, but the friend distribution may also be obtained by other statistical processes. In the generation (calculation) of the friend distribution, weighting may be performed on each place of residence information based on the place of residence information.

[0030] The activity area estimation unit 105 estimates the activity area of the target user based on the generated post distribution and the generated friend distribution. The activity area estimation unit 105 generates the activity area distribution of the target user by overlapping the post distribution and the friend distribution. The granularity level of the generated activity area distribution is the same as the granularity of the post distribution and the friend distribution (or either of them), but it may also be a different granularity. The activity area estimation unit 105 estimates the activity area according to the overlap (the amount of overlap) between the post distribution and the friend distribution. The overlap of the distributions is represented by the scores of the post distribution and the friend distribution respectively obtained by the kernel density estimation function. That is, the activity area is estimated based on the score of the post distribution obtained by the kernel density estimation function and the score of the friend distribution obtained by the kernel density estimation function. The activity area estimation unit 105 estimates the activity area based on a predetermined calculation result of the scores of the post distribution and the friend distribution respectively obtained by the kernel density estimation function. For example, take the product of the score of the post distribution and the score of the friend distribution, and set the area with the highest score as the activity area. Note that it is not limited to the product, and addition, subtraction, etc. may also be used. By multiplying or adding the scores of the post distribution and the friend distribution, the daily activity area of the target user can be estimated. By subtracting the score of the friend distribution from the score of the post distribution, the non-daily activity area can be estimated. The activity area estimation unit 105 may set the area where the obtained score is equal to or higher than a predetermined value as the activity area, or may set the area of the top N items (such as the top 5 items) as the activity area.

[0031] The activity area output unit 106 outputs the estimated activity area. The activity area output unit 106 may be used as a display device to display the activity area in a predetermined format through a GUI (Graphical User Interface). The post distribution and the friend distribution may be displayed, and the area where the distributions overlap may be highlighted. For example, the scores of each activity area may be displayed in the form of a heat map. Also, it may be output to the outside as a file in a predetermined format. For example, the scores of each activity area may be output in a list format, and only a predetermined number of items may be output.

[0032] Figure 3 shows an example of the operation (activity area estimation method) of the activity area estimation device according to the present embodiment. As shown in Figure 3, first, the activity area estimation device 100 identifies the target account of the target user (S101). The post information acquisition unit 101 receives an input of information regarding the target account and identifies the target account based on the input information. The account may be identified by inputting the account ID (identification information) of the target account, or the account may be identified by searching on social media or the Internet from the input name, keyword, etc.

[0033] Subsequently, the activity area estimation device 100 acquires the post information of the target account (S102). The post information acquisition unit 101 accesses the server or database of the social media system 200 and acquires the publicly available account information of the target account. For example, the account information of the target account is acquired within the range possible by the API of the social media service, etc. The post information acquisition unit 101 acquires all the post information included in the account information of the target account.

[0034] Subsequently, the activity area estimation device 100 extracts the posting location and posting date / time of the posting information (S103). The posting information acquisition unit 101 extracts the posting location and posting date / time from all the posting information of the target account. Note that not only all the posting information but also the posting location and posting date / time may be extracted from some of the posting information. For example, posting information older than a predetermined date / time may be excluded from the extraction target, or when there are two pieces of posting information with the same posting content, one of the posting information may be excluded from the extraction target. When a GEO tag is attached to the posting image, the posting information acquisition unit 101 acquires the posting location (location information) from the GEO tag. When a GEO tag is not attached to the posting image, the posting information acquisition unit 101 may analyze the image captured in the posting image and acquire the posting location from a building, scenery, etc. whose location can be specified. When the location information cannot be acquired from the posting image, the posting information acquisition unit 101 may perform natural language processing on the text of the posting and acquire the posting location from words that can specify the location. When the posting location cannot be acquired from the posting information, the posting information acquisition unit 101 may exclude the posting information from the information for generating the posting distribution. Also, the posting information acquisition unit 101 acquires the date / time attached to the posting information as the posting date / time.

[0035] Subsequently, the activity area estimation device 100 generates a posting distribution of the target account (S104). The posting distribution generation unit 102 generates a posting distribution based on the posting locations and posting date / times of the extracted plurality of posting information. In this example, the posting distribution generation unit 102 uses a kernel density estimation function to obtain the posting distribution p(L p ) by the following formula (1). The posting distribution p(L p ) is a set of kernel density estimation values (scores) of the posting information in each distribution area.

Equation

[0036] In formula (1), l p is a set of posting locations, h p is the posting bandwidth, w p is the posting weight, K pIt is a kernel function for posting. The bandwidth is a parameter indicating the influence range of each sample in kernel density estimation. The posting bandwidth is a predetermined value for the posting distribution and may be set in advance or may be a value obtained by learning from a plurality of posting locations in advance. The posting bandwidth may be changed according to the output activity area (estimation result).

[0037] Figure 4 shows an image of the posting distribution obtained by kernel density estimation. As shown in Figure 4, the posting locations of each posting information are plotted on a two-dimensional coordinate of latitude and longitude, and a distribution showing the influence range (for example, a circular shape of a normal distribution) of the posting bandwidth centered on the posting location is obtained. In the influence range of each posting location (sample), the score of the center (posting location) is the largest, and the score decreases as the distance from the center increases. In the example of the figure, the larger the score, the darker the color is shown.

[0038] The posting weight in Equation (1) is the weight of the posting information in the posting distribution based on each posting information. The posting weight indicates the degree of importance of each posting information and sets the magnitude of the score. As an example, the posting weight is a weight based on the posting date and time of the posting information. For example, as shown in Figure 5, the importance of the posting information and the elapsed time are in an inverse proportional relationship, and the importance decreases as time passes. Therefore, the weight is increased (importance is increased) for new posting information and decreased (importance is decreased) for old posting information. By changing the weight in Equation (1) according to the posting date and time, the influence range remains unchanged, but the score can be made larger for new information and smaller for old information.

[0039] On the one hand, after identifying the target account (S101), the activity area estimation device 100 identifies the friend accounts (S105). The friend information acquisition unit 103 identifies the friend accounts that are in a friendship relationship or the like with the target account from the account information of the target account. For example, an account registered as a friend in the account information of the target account is set as a friend account. Also, an account having a relationship such as following or followers of the posts of the target account, an account having post information that quotes the post information of the target account, an account that has given a "like" or the like to the post information of the target account, or an account having an exchange history of messages may be set as a friend account.

[0040] Subsequently, the activity area estimation device 100 acquires the friend information of the friend accounts (S106). Similar to the acquisition of the account information of the target account, the friend information acquisition unit 103 acquires the account information of all friend accounts within the range possible through the API or the like of the social media service from the server or the like of the social media system 200.

[0041] Subsequently, the activity area estimation device 100 extracts the residence information of the friend accounts (S107). The friend information acquisition unit 103 extracts the residence information from the account information of all the acquired friend accounts. The friend information acquisition unit 103 acquires the profile information of the friend's account information and acquires the residence information registered in the profile information. When the residence cannot be obtained from the profile information, the activity bases such as the place of origin, workplace, school, etc. registered in the profile information may be used as the residence information. The posting location may be extracted from the posting information, and the location with the highest frequency of the posting location may be used as the residence information. Also, when the residence information cannot be obtained from the account information of the friend account, the residence of the friend may be estimated from the account information of the friend of the friend (other friend) who is further in a friendship relationship with the friend. For example, the residence of the friend may be estimated based on the distribution of the residences obtained from the account information of the friend of the friend. That is, a friend distribution may be generated based on the residence of the friend specified from the residence of the friend of the friend. When the residence information of the friend account cannot be obtained, the friend information acquisition unit 103 may exclude the information of the friend account from the information for generating the friend distribution.

[0042] Subsequently, the activity area estimation device 100 generates a friend distribution of the friend accounts (S108). The friend distribution generation unit 104 generates a friend distribution based on the residence information of the plurality of extracted friend accounts. In this example, the friend distribution generation unit 104 uses the kernel density estimation function to obtain the friend distribution p(L f ) by the following formula (2), similar to the posting distribution. The friend distribution p(L f ) is a set of kernel density estimation values (scores) of the friend information in each distribution area.

Equation

[0043] In formula (2), l f is the set of the residences of the friends, h f is the bandwidth for friends, w f is the weight for friends, K fIt is a kernel function for friends. The bandwidth for friends is a predetermined value for the friend distribution. Similar to the bandwidth for posts, it may be set in advance or may be a value obtained by learning from the residential locations of multiple friends. The bandwidth for friends may be different from or the same as the bandwidth for posts. The bandwidth for friends may be changed according to the output activity area (estimation result).

[0044] The weight for friends in Equation (2) is the weight of the friend information (residential location) in the friend distribution based on each friend information (account information). The weight for friends indicates the degree of importance of each friend information and sets the magnitude of the score. As an example, the weight for friends may be a weight based on the time when the target user became friends (established a friendship, had a connection) with the friend. For example, when the date and time when the target user became friends can be obtained, old friend information has a small weight (not highly regarded), and new friends have a large weight (highly regarded). This is because when the target user moves, old friends may live near the original address. Conversely, the weighting may be such that new friends are not highly regarded. For example, if there is a desired city or a city where one wants to live, it is estimated that the person has become friends with people living in that city before moving for information collection in that city. In such a case, older friends may be more highly regarded. As a specific calculation method, the weight value may be set to an initial value (100), for example, and this weight value may be decreased based on the passage of time since the target user became friends with the friend. In a simple example, it may be obtained by a linear function such as weight = ax + b (a is a negative value, x is the number of days passed, b is the initial value of 100). Also, a certain reference date may be set, and a certain weight may be given if the person became friends within x days, and no weight may be given if the person became friends more than x days ago.

[0045] Also, the weight for friends may be a weight based on the frequency of conversation such as the number of mentions or retweets for the target user's account. For example, for friends with a higher frequency of conversation with the target user compared to other friends, increase (attach importance to) the weight. As a specific calculation method, the total number of conversations of the target user may be used as the denominator, and the number of conversations with each friend may be used as the numerator to assign a weight to that friend, or weights may be assigned to friends who have had a certain number of conversations or more, and no weights may be assigned to friends who have had less than a certain number of conversations.

[0046] Furthermore, the weight for friends may be a weight based on the reliability of the friend account. Since there are fake accounts that fabricate information among social media users, if such a fake account is included in friends, it may be possible to not attach importance to the information of that friend and make an estimation. The reliability indicates the degree of reliability of the account, and the higher the reliability value, the higher the reliability. The reliability may be a numerical index obtained by distance. The activity area estimation device 100 may further include a reliability calculation unit (not shown), and the reliability calculation unit may obtain the reliability based on the personal attribute information of the account. For example, the reliability calculation unit acquires the personal attribute information (information such as a profile) of the account to be determined for which the reliability is to be obtained and the personal attribute information of the friend accounts of the account to be determined, and estimates the personal attributes of the account to be determined from the personal attribute information of the friend accounts. If the personal attribute information of the friend account includes the place of residence, the place of residence of the user of the account to be determined is estimated based on the physical distance from the place of residence. Furthermore, the reliability is calculated based on the distance between the acquired personal attribute information (place of residence) of the account to be determined and the estimated personal attribute information (place of residence) of the account to be determined. For example, the reliability (or a value based on the reliability) obtained by the reliability calculation unit is used as the weight for friends.

[0047] Also, the weight for friends may be a weight based on the offline friendship degree of friends. An offline friend is a friend who is also in a friendship relationship (connected) with the target user in the physical space (real world) among the friend accounts that are in a friendship relationship with the target user on social media. Estimation may be performed while attaching more importance to the information of these offline friends than the information of online friends. The offline friendship degree indicates whether an offline friendship relationship is formed in the physical space. The activity area estimation device 100 may further include an offline friend determination unit, and the offline friend determination unit may calculate a score indicating the degree of offline friends for each friend account of the target user. Specific examples of the offline friend determination unit and the calculation method of the offline friendship degree will be described in the embodiments described later. For example, the offline friendship degree (or a value based on the offline friendship degree) obtained by the offline friend determination unit is used as the weight for friends.

[0048] FIG. 6 shows an image of the friend distribution obtained by kernel density estimation. As shown in FIG. 6, similar to the post distribution, the residence locations of each friend are plotted on a two-dimensional coordinate of latitude and longitude, and a distribution is obtained that shows the influence range (for example, a circular shape of a normal distribution) of the friend band width centered on the residence location of the friend.

[0049] Following the generation of the post distribution and the generation of the friend distribution, the activity area estimation device 100 generates an activity area distribution of the target user (S109). The activity area estimation unit 105 generates an activity area distribution of the target user by superimposing the post distribution and the friend distribution in the same area (space). For example, the activity area estimation unit 105 obtains the product of the post distribution and the friend distribution obtained from the above equations (1) and (2) as in the following equations (3) and (4), thereby estimating the activity area l t (estimated activity area).

Equation

[0050] In Equation (3), L is l f and l pIt is a set. As shown in formula (4), the score p(L) of each distribution area is proportional to the score of the posting distribution and the score of the friend distribution. As shown in formula (3), the area with the highest score p(L) is estimated as the activity area.

[0051] Figure 7 shows an image in which the posting distribution and the friend distribution are superimposed on the same space (coordinates). As shown in Figure 7, the influence range of each location in the posting distribution and the influence range of each location in the friend distribution are superimposed. The area where the residence of friends and the distribution of posting locations overlap is the activity area, and the area where the amount of overlap is larger (the darker area) is regarded as the activity area.

[0052] Subsequently, the activity area estimation device 100 outputs the generated activity area distribution (S110). The activity area output unit 106 displays the generated activity area distribution in a predetermined format or the like. Figure 8 shows an example of the display of the activity area distribution. As shown in Figure 8, for example, the activity area distribution is displayed by a heat map. In the heat map, on a map (world map, map of Japan, map of a region, etc.), a distribution of colors and densities corresponding to the scores of each area is displayed.

[0053] As described above, in the present embodiment, a place with a stronger trace of activities such as a place with a certain relationship is regarded as the activity area. Specifically, a distribution based on friend information (residence) and a distribution based on posting information (posting location) are generated in parallel at the same time, and the activity area distribution of the target user is generated by superimposing them.

[0054] Comparing Non-Patent Documents 1 and 2 with this embodiment, in Non-Patent Documents 1 and 2, a large amount of data is required for position estimation. That is, in Non-Patent Documents 1 and 2, it is necessary to prepare a large amount of data such as a learning dataset of the positional relationship between a related location and the user to be estimated, posts of friends directly used for estimation, and data that requires a high cost for collecting friends' friend information, etc. However, since social media operating companies have restrictions on data collection (such as the number of data that can be obtained within a certain period), the method using a large amount of data incurs a high cost for data collection. In contrast, in this embodiment, by using an estimation method that does not require prior model preparation, it is not necessary to prepare a large amount of data. Specifically, kernel density estimation that does not require parameter learning using a large amount of data is utilized. Also, by limiting the information used for estimation to the residential areas of the target user's friends and the posting locations of the target user himself / herself, the data collection cost can be reduced. Furthermore, the collection cost can be reduced both during learning and estimation.

[0055] Also, in Non-Patent Document 1, when the target user has few friends or the information obtained from friends is scarce, the position cannot be estimated with high accuracy. That is, in Non-Patent Document 1, in order to discriminate offline friends from the target user's friends and estimate the activity area of the target user by emphasizing the information of the discriminated offline friends, it becomes difficult to estimate when the number of friends and the information of friends are scarce. In contrast, in this embodiment, the activity area of the target user can be estimated based on two types of information. Specifically, the information used for estimation is the residential areas of the target user's friends and the posting locations of the target user himself / herself. Thereby, it is possible to estimate the activity area even for a target user who can obtain only one of the two types of information. Also, by narrowing down to the above two types of information, it is possible to suppress the collection cost more than in Non-Patent Document 1.

[0056] (Embodiment 2) Hereinafter, Embodiment 2 will be described with reference to the drawings. In this embodiment, an example of filtering posting information and friend information in the activity area estimation device of Embodiment 1 will be described.

[0057] Figure 9 shows a configuration example of the activity area estimation device according to the present embodiment. As shown in Figure 9, the activity area estimation device 100 according to the present embodiment includes a posting information filter unit 107 and a friend information filter unit 108 in addition to the configuration of the first embodiment.

[0058] The posting information filter unit 107 filters the posting information of the target account acquired by the posting information acquisition unit 101 under a predetermined condition. The posting information filter unit 107 is a selection unit (first selection unit) that selects the posting information to be used for generating the posting distribution from a plurality of posting information included in the account information of the target user. The posting information filter unit 107 selects the posting information based on the granularity of the posting location, and for example, excludes the posting information whose granularity of the posting location is larger than a predetermined granularity level. As a specific example, the posting information with a granularity at the national or prefectural level, which is larger than the city, town, and village unit, may be excluded, or the posting information with a granularity at the 1 Km×1 Km unit or 100 m×100 m unit, which is larger than the 10 m×10 m unit, may be excluded.

[0059] The friend information filter unit 108 filters the friend information of the friend account acquired by the friend information acquisition unit 103 under a predetermined condition. The friend information filter unit 108 is a selection unit (second selection unit) that selects the residence information (activity base information) to be used for generating the friend distribution from a plurality of residence information included in the account information of the friend. Similar to the posting information, the friend information filter unit 108 selects the residence information based on the granularity of the residence information, and for example, excludes the friend information whose granularity of the residence information is larger than a predetermined granularity level.

[0060] FIG. 10 shows an operation example of the activity area estimation device according to the present embodiment. As shown in FIG. 10, after the extraction of the posting location and posting date and time (S103), the posting information filter unit 107 filters the posting information (S111). The posting information filter unit 107 determines the granularity of the posting location of each extracted posting information, and if the granularity of the posting location is larger than a predetermined granularity level, the posting information is excluded from the information for generating the posting distribution. For example, the predetermined granularity level is the granularity level of the posting distribution (or the activity area distribution to be output) to be generated. Subsequently, the posting distribution generation unit 102 generates a posting distribution from the filtered posting information in the same manner as in the first embodiment (S104).

[0061] Note that in this example, the posting information is filtered according to the granularity of the posting location, but filtering may be performed based on other criteria. The posting information may be filtered based on the posting date and time or the like used as the posting weight in the first embodiment. For example, posting information with a posting date and time older than a predetermined date and time may be excluded.

[0062] Also, in this example, the granularity of the posting location is used as a filtering criterion, but the granularity of the posting location may be used as the posting weight in the first embodiment. That is, in the above formula (1), the posting weight (w p ) may be set as a weight based on the granularity level of the posting location, and a posting distribution may be generated. For example, a more detailed distribution can be generated as the granularity of the posting location is smaller. Therefore, the weight may be increased as the granularity of the posting location is smaller, and the weight may be decreased as the granularity of the posting location is larger.

[0063] On the other hand, after extracting the residence information of friends (S107), the friend information filter unit 108 filters the friend information (S112). Similar to the post information, the friend information filter unit 108 determines the granularity of the residence information of each extracted friend. If the granularity of the friend's residence information is larger than a predetermined granularity level, the friend information is excluded from the information for generating the friend distribution. For example, the predetermined granularity level is the granularity level of the friend distribution (or the activity area distribution to be output) to be generated. Subsequently, the friend distribution generation unit 104 generates a friend distribution based on the filtered friend information in the same manner as in the first embodiment (S108).

[0064] Note that, similar to the post information, filtering may be performed not only based on the granularity of the residence information but also according to other criteria. The friend information may be filtered based on the time when the friendship was established, the conversation frequency, the reliability of the friend account, the offline friend degree of the friend, etc., which were used as the weights for friends in the first embodiment. For example, friend information where the time when the friendship was established with the target user is older (or newer) than a predetermined date and time, friend information where the number of conversations with the target user is less than a predetermined number, friend information where the reliability of the friend account is less than a predetermined value, friend information where the offline friend degree is less than a predetermined value, etc. may be excluded.

[0065] Also, similar to the post information, not only the granularity of the residence information may be used as the filtering criterion, but it may also be used as the weight for friends in the first embodiment. That is, in the above formula (2) of the first embodiment, the weight for friends (w f ) may be set as a weight based on the granularity level of the residence information (activity base) of the friend, and a friend distribution may be generated. For example, similar to the post information, the smaller the granularity of the residence information, the larger the weight, and the larger the granularity of the residence information, the smaller the weight may be set.

[0066] As described above, in this embodiment, the post information for generating the post distribution and the friend information for generating the friend distribution are filtered based on their respective information. As a result, a distribution can be generated based on information at a predetermined granularity level, and thus a distribution with a desired accuracy can be obtained.

[0067] (Embodiment 3) Next, Embodiment 3 will be described with reference to the drawings. In this embodiment, an example of weighting the posting distribution and the friend distribution to be superimposed in the activity area estimation device of Embodiment 1 or 2 will be described.

[0068] FIG. 11 shows a configuration example of the activity area estimation device according to this embodiment. As shown in FIG. 11, the activity area estimation device 100 according to this embodiment includes a weighting unit 109 in addition to the configuration of Embodiment 1. The weighting unit 109 performs weighting (superimposition weighting) on the posting distribution and the friend distribution to be superimposed. For example, weighting may be performed on the friend distribution and the posting distribution according to the number of friend information (sample size) in the friend distribution and the number of posting information (sample size) in the posting distribution, and weighting may be performed according to the difference between the number of friend information and the number of posting information. Also, weighting may be performed on either the friend distribution or the posting distribution. The activity area estimation unit 105 estimates the activity area of the target user based on the weighting of the posting distribution and the friend distribution (or either of them).

[0069] FIG. 12 shows an operation example of the activity area estimation device according to this embodiment. As shown in FIG. 12, after generating the posting distribution (S104) and the friend distribution (S108), the weighting unit 109 performs superimposition weighting on the friend distribution and the posting distribution (S113). The weighting unit 109 counts the number of posting information (posting location) in the generated posting distribution and the number of friend information (residence location) in the generated friend distribution, obtains the difference between the number of posting information and the number of friend information, and performs weighting on the posting distribution and the friend distribution according to the obtained difference. For example, if there is a large difference between the number of posting information and the number of friend information, there is a risk that either piece of information may be overemphasized, so the balance between the number of posting information and the number of friend information may be adjusted. For example, when the number of friends is 100 and the number of posts is 200, the friend distribution and the posting distribution may be superimposed at a ratio of 2 to 1.

[0070] Subsequently, the activity area estimation unit 105 generates an activity area distribution by superimposing the weighted friend distribution and the posting distribution (S109). For example, as in the following equation (5), the score p(L) is obtained by multiplying the weight WF of the friend distribution and the weight WP of the posting distribution by their respective distributions. [Number]

[0071] As described above, in this embodiment, when superimposing the friend distribution and the posting distribution, each distribution is weighted. As a result, it is possible to estimate the activity area of the target user while emphasizing either the friend distribution or the posting distribution. For example, by weighting based on the number of friends and the number of posts, the activity area can be estimated in a well-balanced manner.

[0072] (Embodiment 4) Hereinafter, Embodiment 4 will be described with reference to the drawings. In this embodiment, as another example of the weighting of the superimposition in Embodiment 3, an example of weighting the distribution of online friends and the distribution of offline friends will be described.

[0073] FIG. 13 shows a configuration example of the activity area estimation device according to this embodiment. As shown in FIG. 13, the activity area estimation device 100 according to this embodiment includes an offline friend determination unit 110 in addition to the configuration of Embodiment 3. The offline friend determination unit 110 discriminates offline friends who are in a friendship relationship with the target user in the physical space (real world) from among the friend accounts that are in a friendship relationship with the target user on the social media. That is, from the friends of the target user, offline friends and online friends other than offline friends are discriminated. The activity area estimation unit 105 estimates the activity area of the target user based on the posting distribution, the friend distribution of offline friends, and the friend distribution of online friends. Also, the activity area is estimated based on the weighting of the friend distribution of offline friends and the friend distribution of online friends.

[0074] FIG. 14 shows an operation example of the activity area estimation device according to the present embodiment. As shown in FIG. 14, after the extraction of the residence of a friend (S107), the offline friend determination unit 110 determines offline friends (S114). The offline friend determination unit 110 determines, based on the account information of the acquired friend accounts, whether each friend having a friend account is also a friend with the target user in the physical space or not a friend in the physical space. The offline friend determination unit 110 obtains the offline friend degree of the friend account and determines an offline friend or an online friend based on the offline friend degree. The offline friend determination unit 110 calculates a score indicating the degree of offline friends for each friend account of the target user. For example, when the score exceeds a certain threshold value, the offline friend degree is set to a value indicating that it is an offline friend (for example, "1"), and when the score is equal to or less than the threshold value, the offline friend degree is set to a value indicating that it is not an offline friend (for example, "0"). The threshold value is arbitrarily set by the user of the activity area estimation device 100, for example.

[0075] The offline friend determination unit 110 may determine whether the friend account is a local account related to a specific area. For example, a local account is an account of a social media that is operated targeting a specific place or area among social media accounts. Examples of local accounts include accounts operated by regionally focused companies such as local newspapers, local governments, and privately run restaurants. The offline friend determination unit 110 may calculate the offline friend degree of a friend based on the determination result of whether the friend account is a local account. For example, the offline friend determination unit 110 refers to the friend information (profile information and posting information) of the friend account, calculates a score according to the presence or absence of information indicating that the account is operated targeting a specific place or area and the excess of such information, and determines whether the friend account is a local account.

[0076] Also, when the offline friend determination unit 110 determines that it is unknown whether a friend account is a local account, it may further refer to the friend information of the friend account to determine whether the friend account is a local account. For example, the offline friend degree of the target user's friend account may be calculated based on whether the further friend's account of the friend account is a local account. In addition, the methods described in Non-Patent Document 1 may be used to distinguish offline friends from online friends.

[0077] The friend distribution generation unit 104 generates the friend distribution of the determined offline friends and the friend distribution of the online friends (S108). Similar to Embodiment 1, the friend distribution generation unit 104 generates the friend distribution of the offline friends based on the residence information of the offline friends, and generates the friend distribution of the online friends based on the residence information of the online friends.

[0078] Subsequently, the weighting unit 109 performs weighting on the generated friend distribution of the offline friends and the generated friend distribution of the online friends (S113). For example, offline friends are more important than online friends with respect to the activity area of the target user. Therefore, weighting is performed so that the friend distribution of the offline friends is more emphasized than the friend distribution of the online friends.

[0079] Subsequently, the activity area estimation unit 105 overlays the weighted friend distribution of the offline friends and the friend distribution of the online friends with the post distribution to generate an activity area distribution (S109). Note that an activity area distribution may be generated by overlaying only the friend distribution of the offline friends and the post distribution. For example, as in the following formula (6), the weight WF off of the friend distribution of the offline friends and the weight WF on of the friend distribution of the online friends are multiplied by their respective distributions, and the product is taken with the post distribution to obtain the score p(L). Note that the friend weight in this case preferably does not include the weight based on the offline friend degree.

Equation

[0080] As described above, in the present embodiment, the friend distribution is divided into a distribution of only offline friends and a distribution of only online friends, and the distribution of offline friends is weighted when the post distributions are superimposed. Thereby, it is possible to estimate the activity area of the target user while emphasizing the friend distribution of offline friends.

[0081] Note that the present disclosure is not limited to the above-described embodiment, and can be appropriately modified without departing from the gist. For example, when acquiring the location information (residence) of the friend account of the target user, if the location information cannot be acquired from the friend account or the location information is old, etc., the method for generating the activity area distribution of the target account in the above-described embodiment may be used to estimate the location of the friend account. Thereby, even when the friend account of the target user does not include location information, the location of the friend account can be estimated.

[0082] Each configuration in the above-described embodiment is constituted by hardware or software, or both, and may be constituted by one piece of hardware or software, or may be constituted by a plurality of pieces of hardware or software. Each device and each function (processing) may be realized by a computer 20 having a processor 21 such as a CPU (Central Processing Unit) and a memory 22 which is a storage device, as shown in FIG. 15. For example, a program for performing the method (estimation method) in the embodiment may be stored in the memory 22, and each function may be realized by the processor 21 executing the program stored in the memory 22.

[0083] These programs can be stored using various types of non-transitory computer readable media and supplied to a computer. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer readable media include magnetic recording media (such as flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (such as magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, semiconductor memories (such as mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (random access memory)). Also, the programs may be supplied to a computer by various types of transitory computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. Transitory computer readable media can supply the programs to a computer via wired communication channels such as electric wires and optical fibers, or wireless communication channels.

[0084] Some or all of the above embodiments can also be described as follows, but are not limited thereto.

[0085] (Appendix 1) A first position distribution generation unit that generates a first position distribution of the target user based on account information of the target user in social media; A second position distribution generation unit that generates a second position distribution of the friend based on account information of a friend related to the target user in the social media; An estimation unit that estimates the activity position of the target user based on the generated first position distribution and the generated second position distribution; An estimation device comprising the above. (Supplementary Note 2) The estimation unit estimates the activity position of the target user according to the overlap between the first position distribution and the second position distribution. The estimation device according to Supplementary Note 1. (Supplementary Note 3) The estimation unit estimates the daily or non-daily activity position of the target user. The estimation device according to Supplementary Note 1 or 2. (Supplementary Note 4) The first position distribution generation unit and the second position distribution generation unit generate the first position distribution and the second position distribution respectively by a non-parametric method. The estimation device according to any one of Supplementary Notes 1 to 3. (Supplementary Note 5) The first position distribution generation unit and the second position distribution generation unit generate the first position distribution and the second position distribution respectively using a kernel density estimation function. The estimation device according to Supplementary Note 4. (Supplementary Note 6) The estimation unit estimates the activity position of the target user based on the score of the first position distribution obtained by the kernel density estimation function and the score of the second position distribution obtained by the kernel density estimation function. The estimation device according to Supplementary Note 5. (Supplementary Note 7) The estimation unit estimates the activity position of the target user based on a predetermined calculation result of the score of the first position distribution and the score of the second position distribution. The estimation device according to Supplementary Note 6. (Supplementary Note 8) The first position distribution generation unit generates the first position distribution based on the posting information included in the account information of the target user. The estimation device according to any one of Supplementary Notes 1 to 7. (Supplementary Note 9) The first position distribution generation unit generates the first position distribution based on the posting location extracted from the posting information. The estimation device according to Supplementary Note 8. (Supplementary Note 10) The first position distribution generation unit extracts the posting location from the image or text included in the posting information. The estimation device according to Supplementary Note 9. (Supplementary Note 11) The first position distribution generation unit weights the posting information in the first position distribution based on the posting information. The estimation device according to any one of Supplementary Notes 8 to 10. (Supplementary Note 12) The first position distribution generation unit performs the weighting based on the posting date and time of the posting information. The estimation device according to Supplementary Note 11. (Supplementary Note 13) The first position distribution generation unit performs the weighting based on the granularity of the posting location of the posting information. The estimation device according to Supplementary Note 11 or 12. (Supplementary Note 14) It includes a first selection unit that selects the posting information to be used for generating the first position distribution from a plurality of posting information included in the account information of the target user. The estimation device according to any one of Supplementary Notes 8 to 13. (Supplementary Note 15) The first selection unit selects the posting information based on the posting date and time of the posting information. The estimation device according to Supplementary Note 14. (Supplementary Note 16) The first selection unit selects the posting information based on the granularity of the posting location of the posting information. The estimation device according to Supplementary Note 14 or 15. (Supplementary Note 17) The second position distribution generation unit generates the second position distribution based on the activity base information included in the account information of the friend. The estimation device according to any one of Supplementary Notes 1 to 16. (Supplementary Note 18) The second position distribution generation unit generates the second position distribution based on the place of residence included in the profile information of the account information of the friend. The estimation device according to Supplementary Note 17. (Appendix 19) The second location distribution generation unit generates the second location distribution based on the activity base information of the account information of other friends related to the friend. The estimation device according to Appendix 17 or 18. (Appendix 20) The second location distribution generation unit weights the activity base information in the second location distribution based on the account information of the friend. The estimation device according to any one of Appendices 17 to 19. (Appendix 21) The second location distribution generation unit performs the weighting based on any one of the time when the friend became a friend with the target user, the conversation frequency between the friend and the target user, the reliability of the friend's account, the offline friend degree of the friend, and the granularity of the activity base information. The estimation device according to Appendix 20. (Appendix 22) It includes a second selection unit that selects the activity base information to be used for generating the second location distribution from the plurality of activity base information included in the account information of the friend. The estimation device according to any one of Appendices 17 to 21. (Appendix 23) The second selection unit selects the activity base information based on any one of the time when the friend became a friend with the target user, the conversation frequency between the friend and the target user, the reliability of the friend's account, the offline friend degree of the friend, and the granularity of the activity base information. The estimation device according to Appendix 22. (Appendix 24) The estimation unit estimates the activity location based on the weighting of the first location distribution or the second location distribution. The estimation device according to any one of Appendices 1 to 23. (Appendix 25) The estimation unit estimates the activity location based on the weighting according to the number of samples of the first location distribution and the number of samples of the second location distribution. The estimation device according to Appendix 24. (Appendix 26) A discriminator that discriminates offline friends with a friendship relationship in a physical space from among a plurality of friends related to the target user is provided. The second position distribution generation unit generates a position distribution of the offline friends and a position distribution of online friends other than the offline friends. The estimation unit estimates the activity position of the target user based on the first position distribution, the position distribution of the offline friends, and the position distribution of the online friends. The estimation device according to any one of Appendices 1 to 25. (Appendix 27) The estimation unit estimates the activity position based on weighting of the position distribution of the offline friends and the position distribution of the online friends. The estimation device according to Appendix 26. (Appendix 28) An output unit that outputs the estimated activity position in a heat map format or a list format of a color corresponding to the first position distribution and the second position distribution is provided. The estimation device according to any one of Appendices 1 to 27. (Appendix 29) Based on the account information of the target user in the social media, a first position distribution of the target user is generated. Based on the account information of friends related to the target user in the social media, a second position distribution of the friends is generated. Based on the generated first position distribution and the generated second position distribution, the activity position of the target user is estimated. Estimation method. (Appendix 30) Based on the account information of the target user in the social media, a first position distribution of the target user is generated. Based on the account information of friends related to the target user in the social media, a second position distribution of the friends is generated. Based on the generated first position distribution and the generated second position distribution, the activity position of the target user is estimated. An estimation program for causing a computer to execute processing.

Explanation of Signs

[0086] 1 Activity area estimation system 10 Estimation device 11 First position distribution generation unit 12 Second position distribution generation unit 13 Estimation unit 20 Computer 21 Processor 22 Memory 100 Activity area estimation device 101 Posting information acquisition unit 102 Posting distribution generation unit 103 Friend information acquisition unit 104 Friend distribution generation unit 105 Activity area estimation unit 106 Activity area output unit 107 Posting information filter unit 108 Friend information filter unit 109 Weighting unit 110 Offline friend discrimination unit 200 Social media system

Claims

1. A first position distribution generation unit that generates a first position distribution of the target user based on the account information of the target user in social media; A second position distribution generation unit that generates a second position distribution of the friend based on the account information of the friend related to the target user in the social media; An estimation unit that estimates the activity position of the target user based on the generated first position distribution and the generated second position distribution; A discrimination unit that discriminates offline friends having a friendship relationship in the physical space from among a plurality of friends related to the target user; Comprising: The second position distribution generation unit generates a position distribution of the offline friend and a position distribution of an online friend other than the offline friend; The estimation unit estimates the activity position of the target user based on the overlap between the first position distribution, the position distribution of the offline friend, and the position distribution of the online friend, and the weights of the position distribution of the offline friend and the position distribution of the online friend, and the weight of the position distribution of the offline friend is larger than the weight of the position distribution of the online friend; Estimation device.

2. The estimation unit estimates the daily or non-daily activity position of the target user. The estimation device according to Claim 1.

3. The first position distribution generation unit and the second position distribution generation unit generate the first position distribution and the second position distribution by non-parametric methods respectively. The estimation device according to Claim 1 or 2.

4. The first position distribution generation unit and the second position distribution generation unit generate the first position distribution and the second position distribution by using a kernel density estimation function respectively. The estimation device according to Claim 3.

5. The estimation unit estimates the activity position of the target user based on the score of the first position distribution obtained by the kernel density estimation function and the score of the second position distribution obtained by the kernel density estimation function. The estimation device according to Claim 4.

6. The first position distribution generation unit generates the first position distribution based on the posting information included in the account information of the target user. The estimation device according to any one of Claims 1 to 5.

7. The second position distribution generation unit generates the second position distribution based on the activity base information included in the account information of the friend. The estimation device according to any one of claims 1 to 6.

8. An estimation method executed by an estimation device, generating a first position distribution of the target user based on account information of the target user in social media; generating a second position distribution of the friend based on account information of a friend related to the target user in the social media; estimating an activity position of the target user based on the generated first position distribution and the generated second position distribution; identifying an offline friend having a friendship relationship in the physical space from among a plurality of friends related to the target user; comprising generating the second position distribution includes generating a position distribution of the offline friend and a position distribution of an online friend other than the offline friend; estimating the activity position includes estimating the activity position of the target user based on an overlap between the first position distribution, the position distribution of the offline friend, and the position distribution of the online friend, and a weight of the position distribution of the offline friend and a weight of the position distribution of the online friend, and the weight of the position distribution of the offline friend is greater than the weight of the position distribution of the online friend; Estimation method.

9. generating a first position distribution of the target user based on account information of the target user in social media; generating a second position distribution of the friend based on account information of a friend related to the target user in the social media; estimating an activity position of the target user based on the generated first position distribution and the generated second position distribution; identifying an offline friend having a friendship relationship in the physical space from among a plurality of friends related to the target user; comprising generating the second position distribution includes generating a position distribution of the offline friend and a position distribution of an online friend other than the offline friend; Estimating the activity location includes estimating the activity location of the target user based on the overlap between the first location distribution, the location distribution of offline friends, and the location distribution of online friends, and the weights of the location distributions of offline friends and online friends, wherein the weight of the location distribution of offline friends is greater than the weight of the location distribution of online friends. An estimation program for causing a computer to execute processing.

Images